Skip to main content
    QHSEtech
    Updated
    All glossary terms
    Content date: · No individual check logged· AI use

    Statistical Process ControlSPC

    Also known as: Control charting, Shewhart control charts, Statistical process monitoring

    Definition

    The use of statistical methods, chiefly control charts, to tell a process's routine variation from signals that a special cause needs finding and fixing.

    Editorial definition: The QHSE Standard, updated 3 October 2026 · https://qhsetech.com/glossary/statistical-process-control

    Scope: Control limits come from the process's own data and are not specification limits, so a process can be in statistical control and still miss its specification. The wording here is from the NIST/SEMATECH e-Handbook; ISO 11462-1:2026 was not read.

    Example for your evaluation: Follow one chart signal from the subgroup that raised it, through the out-of-control action plan and the cause found, to the decision on whether the limits change.

    Sources checked: . Examples are editorial illustrations.

    What the NIST/SEMATECH e-Handbook says about SPC, control limits and causes of variation

    The main source here is the NIST/SEMATECH e-Handbook of Statistical Methods, published by the US National Institute of Standards and Technology. Its glossary describes statistical process control as turning process data into information with statistical techniques, so that process performance can be documented, corrected and improved. A note says the glossary was adapted from one donated by an Advanced Micro Devices engineer and may hold terms the handbook does not use. Section 6.1.2 lists the monitoring tools, control charts among them, and separates SPC, which watches a process while it runs, from statistical quality control, which samples finished lots to decide on acceptance.

    The glossary separates two kinds of variation. Common causes act on every value of the output, are typically numerous and individually small, and are difficult to eliminate. A special cause, which it also calls an assignable cause, is big, sporadic or unpredictable and touches only some of the values. Control limits are calculated from process data to mark what common causes alone would produce, so a point beyond one suggests a special cause at work. A process under common causes only is in statistical control; being out of control is a state at a moment, shown by a point outside the limits or a significant trend or pattern inside them. Section 6.1.2 adds two phases: Phase I sets the first limits from historical data, investigating and perhaps discarding points outside them, and Phase II monitors in real time against the result.

    Section 6.3.2 gives each kind of limit its own job: control limits show whether the process is in statistical control, and specification limits whether the product will work as intended. It quotes Shewhart's rule of thumb: no fewer than 25 in-control subgroups of four, gathered under the same essential conditions, before control is claimed, and advises changing limits only for a valid, compelling reason, such as 30 further points with no known process change. Among the Western Electric (WECO) rules it lists eight consecutive points on one side of the centre line, and notes that adding those rules raises false alarms from about one in 371 points under the three-sigma rule alone to about one in 92.

    Section 6.1.4 says that when a chart shows an out-of-control condition, the process engineer looks for an assignable cause by following the out-of-control action plan, or OCAP, attached to that chart. Section 6.1.5 adds that in control only means statistically predictable: the process can still run off target or vary too much, and it must be stable before it can be centred or its variation cut. ISO's catalogue lists ISO 11462-1:2026, Implementation of statistical process control (SPC), Part 1, as edition 2, published in August 2026 and replacing the 2001 Part 1 and the 2010 Part 2. Its summary describes guidance on putting an SPC system in place across the supply chain. The standard is sold, and this page did not read it.

    What we could not verify on 22 October 2026: what ISO 11462-1:2026 says on choosing charts, signal rules or when to recalculate limits; and the 1987 study by Champ and Woodall that the handbook cites for its false-alarm figures, which was not opened. This is general information, not certification advice.

    Sources opened :

    What an SPC record holds in software

    • The characteristic charted, its unit, the measurement method and the gauge identifier with its calibration status.
    • The sampling plan: subgroup size, frequency and how each subgroup is formed.
    • The chart type, centre line and control limits, with the data period behind them and the approver.
    • The specification limits, held apart from the control limits and tied to the drawing or specification revision.
    • The signal rules switched on for that chart, such as a point beyond a limit or a run on one side of the centre line.
    • Every value with its time, subgroup and source, and any correction with its reason, its author and the original value.
    • For each signal, the OCAP followed, the cause found, the action taken and a link to any nonconformance or corrective action.

    Control limits, specification limits, and common versus special causes

    Control limits come from the process's own data and describe what it does. Specification limits come from the design, drawing or customer and describe what is required. The NIST handbook keeps the two apart, so a chart signal is not a failed part and a quiet chart is not proof that parts conform. Common and special causes are the glossary's terms for routine and unusual variation; assignable cause is its synonym for special cause. Comparing a stable process's spread with the specification is process capability.

    A hypothetical example: a run of eight high points on a powder-coating chart

    The line, the people and the numbers below are made up to show the records. A powder-coating line has a specification of 60 to 100 µm for film thickness on a steel bracket, and every hour an operator measures 4 panels. In Phase I the quality engineer collects 25 subgroups, or 100 panels over three shifts, finds no point beyond the trial limits and approves a centre line of 80 µm with control limits for the subgroup mean of 74 µm and 86 µm. Two rules are switched on: a mean beyond either limit, and eight consecutive means on one side of the centre line.

    In Phase II, eight consecutive subgroup means on a Tuesday fall between 81 µm and 84 µm. No point is beyond a limit and all 32 panels measured in that run meet the specification, but the run rule fires and opens an OCAP record. The operator first checks the gauge against its reference standard, which reads within tolerance, so the data stand. The cause is a new powder lot that builds faster at the same gun setting. The setting is lowered, the next 10 subgroup means sit between 78 µm and 82 µm, and the record says why the parts were released. The limits stay as they were, because the cause was found and removed.

    Common misunderstandings about statistical process control

    • Points inside the limits do not settle the matter. The NIST glossary also treats a significant trend or pattern within the limits as out of control.
    • In control does not mean acceptable. Section 6.1.5 says a predictable process can still be off target or too variable.
    • Recalculating limits after each alarm empties the chart of meaning. The handbook asks for a valid, compelling reason, such as a known process change.
    • Extra rules have a price. By the handbook's figures, the WECO rules catch drifts sooner but raise false alarms roughly fourfold.
    • A chart is only as reliable as the measurements behind it, so a gauge check belongs early in the response to a signal.

    How statistical process control relates to other glossary terms

    SPC watches a process while it runs. These terms cover the measurements behind it, what it is compared with and what follows a signal.

    • Process capability: the comparison of a stable process's spread with its specification limits.
    • Measurement system analysis: the study of the gauge, which tells whether a signal could be measurement noise.
    • Quality control: the checks and release decisions that chart data often feed.
    • Non-conformance: the record raised when a product, not only a chart point, misses its requirement.
    • CAPA: the process for a cause that a chart signal brings to light.

    Records and evidence that usually exist for statistical process control

    • Approved chart set-ups with limits, the data period behind them and the approver.
    • Raw measurements with times, subgroup identifiers and the gauge used.
    • Calibration records for the gauges behind each chart.
    • Signal and OCAP records with the cause, the action and the closure.
    • Limit recalculation records with the reason and the data used.
    • Links from signals to nonconformance and corrective action records.

    What a software tool needs to capture for statistical process control

    • Charts set up per characteristic with chart type, subgroup size and the chosen signal rules.
    • Control limits and specification limits held as separate fields, never derived from each other.
    • Phase I limit calculation with a record of each excluded point and the reason.
    • Alerts that open an OCAP or response record and route it to a named person.
    • A gauge field linked to calibration status, with a warning when calibration is overdue.
    • Export of raw values, limits and signals for a chosen period.

    Catalog profiles documenting SPC or process-capability records

    The list follows the SPC or process-capability rule used by the statistical process control software guide. It does not show that a product separates control and specification limits, applies your chosen run rules or keeps a history of limit changes, which should be tested with your own data.

    Same rule for every product: tagged “Statistical Process Control (SPC)”, “SPC”, “Manufacturing Quality”, “Quality Inspection” or “Quality Management”, with a sourced SPC or process-capability capability in its record, or one the vendor stated to us, labelled as vendor-stated. Documented and conditional capabilities qualify; check the linked scope before treating a module as included.

    Tekmon pays for a sponsored placement (above); its place in this list follows the same rule as every entry. How lists are ordered

    All 4 matching profiles are shown. Documentation review, not hands-on testing or a claim of compliance.

    Related software

    Frequently asked questions

    What is the difference between common cause and special cause variation?

    In the NIST handbook's glossary, a common cause affects every value, is usually one of many small sources and is hard to remove, while a special cause is large, intermittent or unpredictable and affects only some values. It uses assignable cause as another name for a special cause.

    How much data is needed before control limits can be set?

    Section 6.3.2 of the handbook gives Shewhart's rule of thumb of at least 25 subgroups of four, in control and gathered under the same essential conditions, before claiming statistical control. It warns that limits from little data may not behave as chart theory assumes, and suggests recalculating once at least 30 further points have come in with no known change to the process.

    Which control chart should be used?

    That depends on the data and on what the chart should catch. The handbook's glossary names X-bar and range charts for rational subgroups, most often of two to six readings, individuals charts for single readings, and CUSUM and EWMA charts, which it describes as sensitive to drift. Agree the chart and its rules before choosing software.

    More quality terms

    Related guides